forums

Miami Ranked #27

tdubs_truckernutz said:
“this is an unbiased review of the performance”

This was created by an espn writer with no background in stem let alone math or computer science. And it is absolutely biased in the strictly mathematical sense of the term and colloquial sense.
It uses Recent Recruiting early in the season to determine so+. The model incorporates a program's recruiting rankings from the previous few seasons to gauge the overall talent level of the roster. The recruit rankings are entirely subjective and often wrong. Go see DMoney’s post on our ranking vs performance under right and Diaz


I have a masters in applied mathematics from Johns Hopkins and a masters in computer science focusing on artificial intelligence from Stanford.
Click to expand...
I did not know you were this respectable of a gentleman, tdub truckernutz.
 
Advertisement
We were ranked 12th last week, but dropped 15 spots to 27. So to be clear, this isn’t about preseason projections. It’s because the model graded our defense poorly against Bethune compared to how it expects other teams would have performed.

That can happen when you face weaker opponents and don’t dominate them quite as much as the model predicts, especially early in the season when there isn't a large enough sample size. Personally, I think we’re much better than 27, so as the season goes on and we build a larger sample against stronger opponents (like USF and UF in the coming weeks), I expect we’ll climb back up.
 
CanesNation26 said:
We were ranked 12th last week, but dropped 15 spots to 27. So to be clear, this isn’t about preseason projections. It’s because the model graded our defense poorly against Bethune compared to how it expects other teams would have performed.

That can happen when you face weaker opponents and don’t dominate them quite as much as the model predicts, especially early in the season when there isn't a large enough sample size. Personally, I think we’re much better than 27, so as the season goes on and we build a larger sample against stronger opponents (like USF and UF in the coming weeks), I expect we’ll climb back up.
Click to expand...

Saying "this isn't about preseason projections" when the model still heavily weighs in preseason projections makes little sense.

FSU has had a dominant 1st 2 weeks and is ranked 31 in SP+. Why? Because of where SP+ had them in the pre-season. If it were just about results on the field FSU would be top 5
 
CanesNation26 said:
We were ranked 12th last week, but dropped 15 spots to 27. So to be clear, this isn’t about preseason projections. It’s because the model graded our defense poorly against Bethune compared to how it expects other teams would have performed.

That can happen when you face weaker opponents and don’t dominate them quite as much as the model predicts, especially early in the season when there isn't a large enough sample size. Personally, I think we’re much better than 27, so as the season goes on and we build a larger sample against stronger opponents (like USF and UF in the coming weeks), I expect we’ll climb back up.
Click to expand...

IMG_5473.jpeg
IMG_5472.jpeg


SIAP. But this how these ratings are made. Pretty useless for the first five weeks or so I’d say.
 
TwentySix said:
Before the “it’s predictive, you just don’t understand metrics” crowd jumps in, I’m very aware.

But at some point we need to stop giving credence to predictive metrics that say a team like Alabama is 11 points better on a neutral field than Miami.
Click to expand...
It's done extremely well vs the spread thru 2 weeks.

It's not perfect, but it's a useful data point.
 
supacane said:
The season we opened with Bama , the returning production said we were #1 and that was a sign of a team that was gonna be good. I drank the kool aid. Never again. Notre Dame is ahead of us as well for all that's worth.
Click to expand...
SP+ is still around years later bc of its great track record.
 
Advertisement
TheRichtShow said:
Saying "this isn't about preseason projections" when the model still heavily weighs in preseason projections makes little sense.

FSU has had a dominant 1st 2 weeks and is ranked 31 in SP+. Why? Because of where SP+ had them in the pre-season. If it were just about results on the field FSU would be top 5
Click to expand...
It actually does make sense. You just missed the point. I’m well aware that preseason projections factor into the model, but that’s not the only thing.

We started the season at 14th in the SP+ preseason rankings. After Notre Dame, we moved up to 12. After Bethune, we dropped to 27.

If this were only about preseason projections, we’d still be sitting at 14. What’s changed is the two games we’ve played, and specifically, the model graded our performance against Bethune poorly. That’s the piece you’re overlooking.

Put differently, if it weren’t for preseason projections, we’d be ranked lower.

You are correct that if it weren’t for preseason projections, FSU would be ranked higher.
 
Last edited:
IndayArtHauz said:
In SP+.

Remember, this is an unbiased review of the performance data of each sector of the game, that’s it.

Dan Lanning’s Oregon Ducks are number 1. Our defense is holding us back in this regard, ranking 62nd.

The SP+ is not suggesting that we are not a better team than those above us, just that we are not as efficient or “successful” as some others. I think we’d agree this rating aligns with what we’ve seen on the field through two games.

View attachment 334817
Click to expand...
And adjusted for opponent, etc. Bill beats Vegas most weeks. Dominate USF and UF and Miami will skyrocket
 
CanesNation26 said:
It actually does make sense. You just missed the point. I’m well aware that preseason projections factor into the model, but that’s not the only thing.

We started the season at 14th in the SP+ preseason rankings. After Notre Dame, we moved up to 12. After Bethune, we dropped to 27.

If this were only about preseason projections, we’d still be sitting at 14. What’s changed is the two games we’ve played, and specifically, the model graded our performance against Bethune poorly. That’s the piece you’re overlooking.

Put differently, if it weren’t for preseason projections, we’d be ranked lower.

You are correct that if it weren’t for preseason projections, FSU would be ranked higher.
Click to expand...
And Bill ranks a lot of the FCS in order to adjust the rankings. It's a good system.
 
Analytics are great and EVERY coach uses them.

What they don’t use is a metric that takes into account last season’s numbers until midway through the current season.

The reason Miami is ranked highly on offense and low on defense is because a majority of the numbers being factored are from games in 2024 when Miami had a great offense and a terrible defense.

Much like rankings, S+P plus is worthless until mid season.
 
Why would an adult with an IQ over say 80 develop a model that uses
"last year’s analytics on defense still taking up a disproportionate amount of the inputs"? LOL, more than 1/2 the defensive starters are new and we have a new DC but your model relies heavily on last year's inputs. You'd flunk an elementary modelling course.

That espn model is not just simplistic it is just silly. When you look at the output of your model and it is obvious you model needs work, you don't publish the results. You rework it. Their "20,000 simulations" doesn't mean anything when the model is fundamentally flawed. You could run a million.

If you think we are the 27th best team in the country (now all the way up to 22nd) and worse than Nebraska, Indiana and much worse than ND who we beat there's nothing I can tell ya.
 
Advertisement
NCCanefan said:
Why would an adult with an IQ over say 80 develop a model that uses
"last year’s analytics on defense still taking up a disproportionate amount of the inputs"? LOL, more than 1/2 the defensive starters are new and we have a new DC but your model relies heavily on last year's inputs. You'd flunk an elementary modelling course.

That espn model is not just simplistic it is just silly. When you look at the output of your model and it is obvious you model needs work, you don't publish the results. You rework it. Their "20,000 simulations" doesn't mean anything when the model is fundamentally flawed. You could run a million.

If you think we are the 27th best team in the country (now all the way up to 22nd) and worse than Nebraska, Indiana and much worse than ND who we beat there's nothing I can tell ya.
Click to expand...

It’s not really true that the previous season isn’t predictive of the next season.

On an individual team level, sure.
But at a macro level, you’re probably going to perform pretty close to how you performed the previous season.
College football isn’t really an itch a sketch season to season where we have no idea who the best teams will be.

And you also need some kind of data points. It makes more sense to just not put too much stock in early season analytic models beyond individual games played. Explosive rate in a game. Success rate. Etc.

But if you are going to at least run the models, then you have to use *something* that gives you a sample size worth anything.
 
By virtue of how it's constructed the model is worthless in the early part of the season. Evidence being the nonsensical rankings after the first two games this season

The model needs data as to how teams perform against each other after repeated tries. This data does not exist after only two games.

I've long felt that this type of ranking (FPI, Sp+) shouldn't even be published until 8 games into the season.
 
mossmadness said:
It’s not really true that the previous season isn’t predictive of the next season.

On an individual team level, sure.
But at a macro level, you’re probably going to perform pretty close to how you performed the previous season.
College football isn’t really an itch a sketch season to season where we have no idea who the best teams will be.

And you also need some kind of data points. It makes more sense to just not put too much stock in early season analytic models beyond individual games played. Explosive rate in a game. Success rate. Etc.

But if you are going to at least run the models, then you have to use *something* that gives you a sample size worth anything.
Click to expand...
If your point is that at a macro level results don't vary much year to year, that tells us that GA will still be GA and East TX A&M wil still be East TX A&M. I hope no one spent a lot of money for that information.

The relevant question is in 2025, does your power ranking model represent reality. If you have all new players and coordinators, who cares what you did in 2024? If the claim is the power ratings don't really predict rankings and you know that is what people use them for then you are just engaging in deception and trying to fill up some space on your ESPN pages.
 
NCCanefan said:
If your point is that at a macro level results don't vary much year to year, that tells us that GA will still be GA and East TX A&M wil still be East TX A&M. I hope no one spent a lot of money for that information.

The relevant question is in 2025, does your power ranking model represent reality. If you have all new players and coordinators, who cares what you did in 2024? If the claim is the power ratings don't really predict rankings and you know that is what people use them for then you are just engaging in deception and trying to fill up some space on your ESPN pages.
Click to expand...

Connelly’s model actually does account for new coordinators in its projections.
 
Back
Top